Postdoc: Physics-Driven ML for TEM Subsurface Imaging

Aarhus Universitet; Favrskov Kommune; On-site; DKK 420,000 - 520,000; Full time

Benefits offered by this job

Relocation support

Housing assistance

Childcare support

Job summary

Aarhus University, Department of Electrical and Computer Engineering, invites applications for a two-year postdoctoral position focusing on physics-driven machine learning for ground-based, airborne, and drone TEM data. The project builds on the anEMone framework to develop supervised and self-supervised models that fuse electromagnetic physics with data-driven methods.

The role includes developing models, losses, and real-time capable implementations, collaborating with geophysicists and

Qualifications

• PhD in a related field with strong publication record.
• Experience with deep learning model development (CNNs, UNets, Transformers).
• Strong Python and PyTorch software skills.
• Experience in scientific computing or inverse problems.

Responsibilities

• Develop supervised and self-supervised deep learning models for TEM inversion that incorporate a differentiable TEM forward operator.
• Design physics-based loss functions and training strategies enforcing consistency between measured data, predicted subsurface models and simulated TEM responses.
• Incorporate system geometry, transmitter waveforms, measurement uncertainties and motion information into learning frameworks.
• Train and validate methods using synthetic and field TEM datasets.
• Develop computationally efficient implementations for near-real-time applications.
• Publish high-impact research and contribute to outreach and dissemination.

Skills

Deep learning

Scientific ML

Inverse problems

English communication

Education

PhD in electrical engineering, computer engineering, computer science, applied mathematics, geophysics, physics or a related field

Tools

PyTorch

Python

Job description

Aarhus University, Department of Electrical and Computer Engineering, invites applications for a two-year postdoctoral position focusing on physics-driven machine learning for ground-based, airborne, and drone TEM data. The project builds on the anEMone framework to develop supervised and self-supervised models that fuse electromagnetic physics with data-driven methods.

The role includes developing models, losses, and real-time capable implementations, collaborating with geophysicists and